Key Points
Question
How well does a clinician-based assessment, the electronic, clinician-graded facial function scale (eFACE), agree with an automated facial-measurement algorithm (Emotrics) in analyzing eyelid weight placement for facial palsy?
Findings
This cohort study of 53 patients with unilateral facial palsy before and after eyelid weight placement was conducted using both eFACE and Emotrics. The eFACE scores agreed well with the Emotrics facial measurements of palpebral fissure at rest, with gentle eyelid closure, and with forceful eyelid closure, all of which significantly improved following eyelid weight placement.
Meaning
The eFACE tool agreed well with the automated, objective facial measurements using a machine-learning based algorithm, Emotrics, potentially allowing physicians to rapidly and accurately assess a patient’s regional facial function before and after surgical intervention in a clinical setting.
This cohort study compares the results of a clinician-based assessment tool with an automated facial-measurement algorithm in a preoperative and postoperative analysis of 53 patients with facial palsy who had eyelid weight-placement surgery.
Abstract
Importance
Quantitative assessment of facial function is difficult, and historic grading scales such as House-Brackmann have well-recognized limitations. The electronic, clinician-graded facial function scale (eFACE) allows rapid regional analysis of static, dynamic, and synkinetic facial function in patients with unilateral facial palsy within the course of a clinical encounter, but it relies on clinician assessment. A newly developed, machine-learning algorithm (Emotrics) provides automated, objective facial measurements but lacks clinical input (ie, recognizing laterality of facial palsy or synkinesis).
Objectives
To compare the sensitivity of a clinician-based tool (eFACE) to a well-established intervention for facial palsy (eyelid weight placement) with an automated facial-measurement algorithm (Emotrics).
Design, Setting, and Participants
A retrospective review was conducted of the most recent 53 patients with unilateral facial palsy who received an eyelid weight at the Massachusetts Eye and Ear Infirmary Facial Nerve Center from 2014 to 2017. Preoperative and postoperative photographs were deidentified and randomized. The entire cohort was analyzed by 3 clinicians, as well as by the Emotrics program.
Main Outcomes and Measures
eFACE scores of the palpebral fissure at rest (0, wide; 100, balanced; 200, narrow), with gentle eyelid closure (0, incomplete; 100, complete), and with forceful eyelid closure (0, incomplete; 100, complete) before and after eyelid weight placement were compared with palpebral fissure measurements by Emotrics.
Results
Of the 53 participants, 33 were women, and mean (SD) age was 44.7 (18) years. The mean (SD) eFACE scores and Emotrics measurements (in millimeters) before vs after eyelid weight placement of the palpebral fissure at rest (eFACE, 84.3 [15.9] vs 109.7 [21.4]; Emotrics, 10.3 [2.2] vs 9.1 [1.8]), with gentle eyelid closure (eFACE, 65.9 [28.0] vs 92.1 [15.4]; Emotrics, 4.4 [2.7] vs 1.3 [2.0]), and with forceful eyelid closure (eFACE, 75.1 [28.6] vs 97.0 [10.7]; Emotrics, 3.0 [3.1] vs 0.5 [1.3]) all significantly improved. Subgroup analysis of patients with expected recovery (eg, Bell palsy) (n = 40) demonstrated significant development of ocular synkinesis on eFACE (83.9 [22.7] vs 98.9 [4.4]) after weight placement, which could also explain the improvement in eyelid function. The scores of patients with no expected recovery (n = 13) improved in both eFACE and Emotrics analysis following eyelid weight placement, though results did not reach significance, likely limited by the small subgroup size.
Conclusions and Relevance
The eFACE tool agrees well with automated, objective facial measurements using a machine-learning based algorithm such as Emotrics. The eFACE tool is sensitive to spontaneous recovery and surgical intervention, and may be used for rapid regional facial function assessment from a clinician’s perspective following recovery and/or surgical intervention.
Level of Evidence
4.
Introduction
Clinical outcomes research in facial palsy is limited by lack of objective and quantitative evaluation of facial function.1 Multiple clinician-reported scales are available such as the electronic, clinician-graded facial function scale (eFACE), Sunnybrook facial grading system, and the House-Brackmann grading scale,2,3,4,5 but usage depends on institutional and individual preferences. Quantitative facial measurements have been used frequently in the literature to assess facial function,6,7,8 but their objectivity is potentially hindered by manual measurements of soft-tissue landmarks. A newly developed, open-source machine-learning algorithm (Emotrics) provides automated, objective facial measurements for patients with facial palsy.9 While machine-learning applications are a powerful tool to improve facial analysis and clinical outcomes research in facial palsy, the data continue to require careful clinical interpretation (eg, recognizing laterality of facial palsy or synkinesis).
The goal of this study was to compare the clinician-based analysis (eFACE) with an automated facial measurement algorithm (Emotrics) through an analysis of a well-established intervention for facial palsy. Eyelid weights are frequently placed in patients with lagophthalmos from facial palsy and can greatly aid eye closure, comfort, and lubrication, as well as prevent corneal keratitis and blindness.10,11 Contemporary eyelid weights are thin profile owing to the density of platinum (as opposed to gold weights), placed easily under local anesthesia, and are removable. We hypothesized that the clinician-based analysis as well as the automated facial measurements would detect the improvement in lagophthalmos (decreased palpebral fissure with gentle and forceful eye closure) following eyelid weight placement. After further review of the patients’ histories, it was noted that a substantial proportion of patients had recovering facial palsy, and a small subset eventually had the eyelid weight removed. Given the potentially transient nature of facial palsy in this study, the study cohort was additionally divided into 2 groups: those with expected facial nerve recovery (postviral or other cause of temporary facial palsy), and those with no expectation of recovery.
Methods
Institutional review board approval was obtained from the Massachusetts Eye and Ear Infirmary Human Studies Committee and written informed consent was received from all patients. A retrospective review of the most recent 53 patients from 2014 to 2017 who underwent eyelid weight placement and had preoperative and postoperative photography was completed. The size of the cohort was based on an a priori, 2-tailed power analysis conducted with G*Power software, version 3.1.9.2, as described by Meadows et al12 (n = 42), comparing lagophthalmos (millimeters) before and after eyelid weight placement. The effect size in this study was considered to be large according to Cohen 1988 criteria. With an α = .05, power of 0.80, the projected sample size required was 20 patients. Because there were no available published data of eFACE values following eyelid weight implant, the cohort size was increased to 55 based on a power analysis using eFACE scores in 10 randomly selected patients. Two patients were found to have insufficiently high-quality photodocumentation and were excluded (partial blinking during rest photographs).
Patients who underwent additional periocular surgery, such as lower eyelid tightening or lateral tarsorrhaphy, were excluded. Standard photographs of the patient’s face at rest, with gentle eyelid closure, forceful eyelid closure, and pucker preoperatively and postoperatively were deidentified and labeled using a randomized identification number. The photographs were randomly assorted, and eFACE scores were given by 3 different facial plastic and reconstructive surgeons (J.G., J.T., T.H.). These photographs were also analyzed using Emotrics to measure the palpebral fissure at rest, with gentle eyelid closure, and with forceful eyelid closure.
Data Collection
Demographic data including age, sex, laterality of facial palsy, and cause of facial paralysis were recorded. The patients’ histories were carefully reviewed to divide the study cohort into 2 groups: those with expected facial nerve recovery (postviral or other cause of temporary facial palsy), and those with no expectation of recovery.
Statistical Analysis
A focused regional analysis of the patient cohort was conducted using the following eFACE parameters: palpebral fissure at rest (0, wide; 100, balanced; 200, narrow), with gentle eyelid closure (0, incomplete; 100, complete), with forceful eyelid closure (0, incomplete; 100, complete), and degree of ocular synkinesis (0, severe; 100, absent). The entire cohort’s eFACE scores from the 3 surgeons were averaged for each patient. These scores were then compared preoperatively and postoperatively using a paired t test. Automated palpebral fissure distance (millimeters) at rest, with gentle eyelid closure, and with forceful eyelid closure were compared from preoperative and postoperative photographs using a paired t test. All statistical analysis was conducted with the R Statistical Package, version 3.2.3.
Results
A total of 53 patients were included in the study, 33 women and 20 men, with a mean (SD) age of 44.7 (18) years. Demographic and clinical details are provided in the Table. The most common cause of facial palsy was vestibular schwannoma, followed by facial nerve schwannoma and Ramsay Hunt syndrome.
Table. Demographic Data of Patient Cohort Undergoing Eyelid Weight Placement.
| Characteristic | No. |
|---|---|
| Age, mean (SD), y | 44.7 (18.0) |
| Sex | |
| Male | 20 |
| Female | 33 |
| Laterality | |
| Right | 31 |
| Left | 22 |
| Cause of facial palsy | |
| Parotid malignancy | 6 |
| Acoustic neuroma | 2 |
| Bell palsy | 4 |
| Intracranial tumor | 5 |
| Cavernous brainstem hemangioma | 1 |
| Iatrogenic | 3 |
| Congenital | 1 |
| Facial nerve schwannoma | 6 |
| Geniculate ganglion hemangioma | 1 |
| Glomus tumor | 2 |
| Cholesteatoma | 1 |
| Neurofibromatosis 2 | 1 |
| Benign parotid tumor | 1 |
| Pregnancy-associated Bell palsy | 1 |
| Ramsay-Hunt syndrome | 5 |
| Temporal bone fracture | 4 |
| Trauma | 1 |
| Unclear origin | 1 |
| Vestibular schwannoma | 6 |
| Viral | 1 |
eFACE Analysis of the Whole Cohort
As expected for this well-established intervention for lagophthalmos, all eFACE parameters (mean score [SD]) significantly improved from preoperative values following eyelid weight placement: the palpebral fissure was significantly more balanced at rest (Figure 1A; 84.3 [15.9] vs 109.7 [21.4]; P < .001), closed more fully with gentle effort (Figure 1B; 65.9 [28.0] vs 92.1 [15.4]; P < .001), and closed more fully with forceful effort (Figure 1C; 75.1 [28.6] vs 97.0 [10.7]; P < .001). Ocular synkinesis was assessed when the patient was asked to pucker their lips, and while preoperatively absent (98.9 [4.4]), it was found to significantly increase following eyelid weight placement (83.9 [22.7]) (P < .001).
Figure 1. eFACE Scores of the Whole Cohort Following Eyelid Weight Placement.
A focused regional analysis of the patient cohort was conducted using the eFACE parameters of palpebral fissure at rest (0, wide; 100, balanced; 200, narrow), with gentle eyelid closure (0, incomplete; 100, complete), with forceful eyelid closure (0, incomplete; 100, complete), and ocular synkinesis (0, severe; 100, absent). Black circles represent individual patient scores and blue squares represent the group average. Fifty-three patients were analyzed preoperatively and postoperatively using the eFACE application with a statistically significant improvement in palpebral fissure at rest (A), with gentle eyelid closure (B), and with full eyelid closure (C). Ocular synkinesis was noted to significantly worsen (as expected for those experiencing spontaneous recovery) by the postoperative period (D).
eFACE Analysis of Patients With No Expected Facial Nerve Recovery
Patients with an intact facial nerve can experience synkinetic facial nerve recovery and can develop narrowing of the palpebral fissure, which could also contribute to the improved eyelid closure following eyelid weight placement. Forty patients had potential for synkinetic facial nerve recovery; 20 patients ultimately had their eyelid weights removed. Thirteen patients had no potential for spontaneous recovery and were analyzed independently. Although palpebral fissure at rest was significantly more balanced following eyelid weight placement (83.2 [14.3] vs 106.7 [24.5]; P = .01), and some improvement in gentle and full eyelid closure was noted, no statistically significant difference was found in these latter 2 parameters (Figure 2). Ocular synkinesis was relatively unchanged during the postoperative period, as expected.
Figure 2. eFACE Scores Following Eyelid Weight Placement for Patients With No Expected Facial Nerve Recovery.
A focused regional analysis of the patient cohort was conducted using the eFACE parameters of palpebral fissure at rest (0, wide; 100, balanced; 200, narrow), with gentle eyelid closure (0, incomplete; 100, complete), with forceful eyelid closure (0, incomplete; 100, complete), and ocular synkinesis (0, severe; 100, absent). Black circles represent individual patient scores and blue squares represent the group average. Thirteen patients were analyzed preoperatively and postoperatively using the eFACE application with a statistically significant improvement in palpebral fissure at rest (A); but no significant change in palpebral fissure with gentle eyelid closure (B), full eyelid closure (C), or ocular synkinesis (D).
Automated Facial Measurements of the Whole Cohort
Palpebral fissure distance (mean [SD]; in millimeters) was measured using an automated facial measurement application (Emotrics). Palpebral fissure at rest decreased significantly from its preoperative value following eyelid weight placement (10.3 [2.2] vs 9.1 [1.8]; P < .001), with gentle eyelid closure (4.4 [2.7] vs 1.3 [2.0] P < .001), and with full eyelid closure (3.0 [3.1] vs 0.5 [1.3]; P < .001) (Figure 3).
Figure 3. Automated Measurements of Palpebral Fissure of the Whole Cohort Following Eyelid Weight Placement.
Black circles represent individual patient scores and blue squares represent the group average. Palpebral fissure distance was significantly decreased at rest (A), with gentle eyelid closure (B), and with forceful eyelid closure (C) following eyelid weight placement.
Automated Facial Measurements of Patients With No Expected Facial Nerve Recovery
No significant change in palpebral fissure was noted in the cohort of 13 patients with no expected facial nerve recovery, although some improvement was found (Figure 4).
Figure 4. Automated Measurements of Palpebral Fissure of Patients With No Expected Recovery Following Eyelid Weight Placement.
Black circles represent individual patient scores. No significant changes in palpebral fissure distance were detected in this smaller (n = 13) cohort for at rest (A), with gentle eyelid closure (B), and with forceful eyelid closure (C), although some improvement was found.
Discussion
Facial function is difficult to quantify, and the lack of a universal grading scale has limited attempts to perform clinical outcomes research.4,5,13,14 The eFACE was developed through clinical insights gained from more than 3000 patients with facial palsy over a 15-year period and is a rapid application that can be used on a variety of devices to provide graphic and quantitative data regarding facial function in the clinic setting.3,15 Unlike the House-Brackmann scale, which was adopted in 1984 by the American Academy of Otolaryngology–Head and Neck Surgery to describe recovery following resection of vestibular schwannomas,16 the eFACE can be used to evaluate patients with facial nerve weakness from any cause. While the Sunnybrook grading scale also allows for zonal facial analysis for both the flaccid and synkinetic states, categorical outcomes (0-2 for resting symmetry, 1-5 for dynamic movement, and 0-3 for synkinesis) may oversimplify clinical states. The eFACE provides a continuous data scale (0-100) for each factor, allowing discernment between subtle changes in facial function. Additionally, while the Sunnybrook scale prioritizes the “snarl” facial expression, which is of questionable clinical relevance, the eFACE assesses lower lip depressor function (saying “eee”), which is associated with smile asymmetry, articulation difficulties, and oral incompetence. The eFACE has a high interrater reliability, and provides quantification of key factors associated with facial disfigurement, such as nasolabial-fold asymmetry, lower-lip asymmetry, and palpebral-fissure asymmetry at rest and while smiling.15
The goal of this study was to test whether the eFACE would produce the same results as an objective, automated measurement of palpebral fissure following a simple surgical intervention. Eyelid weights are a well-established, routinely performed surgical intervention for lagophthalmos from facial palsy with a univector, measurable outcome (eyelid closure or palpebral fissure distance). We found that the eFACE and Emotrics analysis both noted significant improvement in eyelid closure (reduction in palpebral fissure) following weight placement, but no significant difference was noted in subgroup analysis of patients with no expected recovery, likely owing to a small sample size. Both the eFACE and Emotrics successfully demonstrated the natural history of recovering facial nerve function: initial severe lagophthalmos that can threaten vision and should be promptly treated with an eyelid weight and careful eye care, followed by recovery and occasional ocular synkinesis that warrants removal of an eyelid weight.
Limitations
All clinician-based grading scales can be faulted for subjectivity, particularly in the setting of surgical intervention. Recent advances in machine-learning algorithms have led to automated facial measurement applications, such as Emotrics, which arguably provide more objective assessment.9 While facial analysis with automated programs like Emotrics are a powerful tool, they are based on high-quality photography and videography and may not be as easily used as the clinician-based scales, which can be performed in real time during a clinical encounter. Additionally, the data must be carefully interpreted to appropriately recognize critical points such as facial palsy laterality, synkinesis with aberrant movements, or the effect of facial reanimation, and any potential confounding factors from imperfect photography or videography conditions. For facial palsy outcomes research purposes, automated and clinician-based facial assessment scales are complementary and, when used concurrently, may offer more robust conclusions of clinical evaluation bolstered by objective facial measurements.
Conclusions
The clinician-based facial grading system (eFACE) and an automated facial measurement algorithm (Emotrics) both demonstrate significant improvement following eyelid weight placement in patients with lagophthalmos from facial palsy. The eFACE tool may be used for rapid regional facial function assessment from a clinician’s perspective following a surgical intervention and agrees well with automated, objective facial measurements using a machine-learning based algorithm.
References
- 1.Bhama P, Gliklich RE, Weinberg JS, Hadlock TA, Lindsay RW. Optimizing total facial nerve patient management for effective clinical outcomes research. JAMA Facial Plast Surg. 2014;16(1):9-14. doi: 10.1001/jamafacial.2013.1382 [DOI] [PubMed] [Google Scholar]
- 2.Rozen SM. Discussion: worldwide testing of the eFACE facial nerve clinican-graded scale. Plast Reconstr Surg. 2017;139(2):499-500e. doi: 10.1097/PRS.0000000000002986 [DOI] [PubMed] [Google Scholar]
- 3.Banks CA, Jowett N, Hadlock TA. Test-retest reliability and agreement between in-person and video assessment of facial mimetic function using the eFACE facial grading system. JAMA Facial Plast Surg. 2017;19(3):206-211. doi: 10.1001/jamafacial.2016.1620 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Kanerva M, Jonsson L, Berg T, et al. Sunnybrook and House-Brackmann systems in 5397 facial gradings. Otolaryngol Head Neck Surg. 2011;144(4):570-574. doi: 10.1177/0194599810397497 [DOI] [PubMed] [Google Scholar]
- 5.Coulson SE, Croxson GR, Adams RD, O’Dwyer NJ. Reliability of the “Sydney,” “Sunnybrook,” and “House Brackmann” facial grading systems to assess voluntary movement and synkinesis after facial nerve paralysis. Otolaryngol Head Neck Surg. 2005;132(4):543-549. doi: 10.1016/j.otohns.2005.01.027 [DOI] [PubMed] [Google Scholar]
- 6.Bhama PK, Weinberg JS, Lindsay RW, Hohman MH, Cheney ML, Hadlock TA. Objective outcomes analysis following microvascular gracilis transfer for facial reanimation: a review of 10 years’ experience. JAMA Facial Plast Surg. 2014;16(2):85-92. doi: 10.1001/jamafacial.2013.2463 [DOI] [PubMed] [Google Scholar]
- 7.Albathi M, Oyer S, Ishii LE, Byrne P, Ishii M, Boahene KO. Early nerve grafting for facial paralysis after cerebellopontine angle tumor resection with preserved facial nerve continuity. JAMA Facial Plast Surg. 2015;18(1):54-60. doi: 10.1001/jamafacial.2015.1558 [DOI] [PubMed] [Google Scholar]
- 8.Snyder-Warwick AK, Fattah AY, Zive L, Halliday W, Borschel GH, Zuker RM. The degree of facial movement following microvascular muscle transfer in pediatric facial reanimation depends on donor motor nerve axonal density. Plast Reconstr Surg. 2015;135(2):370e-381e. doi: 10.1097/PRS.0000000000000860 [DOI] [PubMed] [Google Scholar]
- 9.Guarin DL, Dusseldorp J, Hadlock TA, Jowett N. A machine learning approach for automated facial measurements in facial palsy. JAMA Facial Plast Surg. 2018;20(4):335-337. [DOI] [PubMed] [Google Scholar]
- 10.Bhama P, Bhrany AD. Ocular protection in facial paralysis. Curr Opin Otolaryngol Head Neck Surg. 2013;21(4):353-357. doi: 10.1097/MOO.0b013e32836239cf [DOI] [PubMed] [Google Scholar]
- 11.Silver AL, Lindsay RW, Cheney ML, Hadlock TA. Thin-profile platinum eyelid weighting: a superior option in the paralyzed eye. Plast Reconstr Surg. 2009;123(6):1697-1703. doi: 10.1097/PRS.0b013e3181a65a56 [DOI] [PubMed] [Google Scholar]
- 12.Meadows A, Hall N, Shah-Desai S, Low JL, Manners R. The House-Brackmann system and assessment of corneal risk in facial nerve palsy. Eye (Lond). 2000;14(Pt 3A):353-357. doi: 10.1038/eye.2000.87 [DOI] [PubMed] [Google Scholar]
- 13.Reitzen SD, Babb JS, Lalwani AK. Significance and reliability of the House-Brackmann grading system for regional facial nerve function. Otolaryngol Head Neck Surg. 2009;140(2):154-158. doi: 10.1016/j.otohns.2008.11.021 [DOI] [PubMed] [Google Scholar]
- 14.Fattah AY, Gurusinghe ADR, Gavilan J, et al. ; Sir Charles Bell Society . Facial nerve grading instruments: systematic review of the literature and suggestion for uniformity. Plast Reconstr Surg. 2015;135(2):569-579. doi: 10.1097/PRS.0000000000000905 [DOI] [PubMed] [Google Scholar]
- 15.Banks CA, Bhama PK, Park J, Hadlock CR, Hadlock TA. Clinician-graded electronic facial paralysis assessment: the eFACE. Plast Reconstr Surg. 2015;136(2):223e-230e. doi: 10.1097/PRS.0000000000001447 [DOI] [PubMed] [Google Scholar]
- 16.Brackmann DE, Barrs DM. Assessing recovery of facial function following acoustic neuroma surgery. Otolaryngol Head Neck Surg. 1984;92(1):88-93. doi: 10.1177/019459988409200119 [DOI] [PubMed] [Google Scholar]
